Trace back taxonomic classification to corresponding read IDs

[Salocin (Marc Bendisch)]

Hi all,

Is there a function within QIIME2 to connect taxonomic classification not only to the corresponding feature ID/sequence but to the read IDs of the sample fasta/fastq file?
I am also open for every custom made solution.

Best wishes,
Marc Bendisch

[timanix (Timur Yergaliyev)]

Hi!
I am not sure if it is what you are looking for, but I wrote a script in Python that modifies sequences IDs, representative sequences and taxonomy file by adding last available taxonomy unit to the sequence ID, so all downstream analysis in Qiime2 may be performed with these modified files. Let me know if you need the code

[Salocin (Marc Bendisch)]

Thanks for the offer @timanix !
Basically, I want to be able to see how every read from my sample file was classified. It would be perfect if I could get the whole lineage, but the highest ranking classification should also be sufficient. Sounds like something your script can provide, if I understand you right.

[timanix (Timur Yergaliyev)]

You can check this link, I provided some examples there.
The code may need some adjustments based on the dataset taxonomy

[Salocin (Marc Bendisch)]

Sadly I don't think your script can solve my problem. What I am interested in is something like this:

Exporting and modifying BIOM tables (e.g. adding taxonomy annotations)

Creating a TSV BIOM table The following is based on a post by @gregcaporaso. $ qiime tools export table.qza --output-dir ./ $ biom convert -i feature-table.biom -o feature-table.tsv --to-tsv $ biom head -i feature-table.tsv # Constructed from biom file #OTU ID L1S105 L1S140 L1S208 L1S257 L1S281 4b5eeb300368260019c1fbc7a3c718fc 2222.0 0.0 0.0 0.0 0.0 fe30ff0f71a38a39cf1717ec2be3a2fc 5.0 0.0 0.0 0.0 0.0 d29fe3c70564fc0f69f2c03e0d1e5561 0.0 0.0 0.0 0.0 0.0 868528ca947bc57b69ffdf83e6b73bae 0.0 227…

But instead of a taxonomic annotation to the corresponding Feature ID, I need taxonomic annotation directly to the corresponding Read ID or at least the whole original read existing in the sample fastq file.

[llenzi (Luca)]

Hi @Salocin,

What command/tool do you use to get the feature table?

If you dealing with OTUs obtained by VSEARCH, you may be able to output the file listing in which OTU fell each sequences. So you can do a bit of scripting to merge all the info into a final table.

If you got ASVs, the things is a bit more complicate because I don’t think dada2 and deblur tracks the destiny of each sequences after the denoising, but just pull out the final count for each unique ASV obtained.
A potential way for you to do what you need is to align back all the sequences onto the sequence feature file (vsearch may do that for you), and from the output file figure out the best match for each sequence. Then, again, merge all the info with a bit of scripting.
(The other way I can see is to taxonomy classify all the sequences, but I assume it is not doable of course …)

I hope it make sense, and it is helpful
Cheers

[Salocin (Marc Bendisch)]

Hi @llenzi,

thanks for the help!
In fact there is a way to track individual reads within the dada2 workflow:

https://github.com/benjjneb/dada2/issues/889

[llenzi (Luca)]

Hi @Salocin,
thanks for sharing the solution you found! Definitely a good thing to know!
Given you will have to work on dada2 within R to use this trick, my suggestion is to run R in the qiime2 conda environment you used for the qiime2 analysis.
The only reason is to be sure to use the same version of dada2 in both the cases (qiime2 and R standalone), and so to be able to replicate (by using same settings) the q2-dada2 plug in result.
Cheers
Luca